cfx-fsi_120_lect-06_twoway[宝典].ppt

* * * * * * * * * * * * * * * * * * * * * * If you disable Process Input File and use an unedited input file from Workbench, then ANSYS will just run the mechanical model “as is” and never establish communication with CFX. * * * In this case the ANSYS field solver is converging fine every stagger loop. The oscillating plot for the interface loads shows that at the start of each timestep the loads are not converged, but they reach convergence at the end of the timestep (each spike is a timestep). * * The field solver convergence and interface convergence are plotted in CFX. It’s useful to keep an eye on how many equilibrium iterations ANSYS needs – too many can really slow down the overall solution. It’s not necessary to fully convergence the ANSYS field solver each stagger loop, just like CFX doesn’t need to be converged until the end of the timestep. How to limit the # of equilibrium iterations is shown later. * * * This shows a disconnected workflow in workbench for a transient 2-way FSI case that requires a steady-state solution as the starting point. * * * Offset notes: Imaging 2-way FSI for flow in a pipe using shell elements. The surface used to generate the shell elements will be located on the inner surface of the pipe, since it must physically match to the fluid domain. By default ANSYS assumes shell elements are located at the mid-point of the thickness. For inward pointing normals, Offset Type = Top will locate the nodes on the top layer of the shells. Since forces are applied at nodes, the forces are now applied at a location consistent with the geometry. * * * * * * * * * * * “steady-state” here means transient dynamic with TIMINT,OFF * * * * * * * “the system is over-constrained” is a bit vague, but I don’t want to get into the details (partly because I’m not clear on the details!). Essentially by neglecting the total energy terms the fluid cannot respond to large pressure waves “correctly”. It should naturally include some small enthalpy